Locations and year

There are 3 years concerned by this study: 2011, 2013, and 2015. There are 5 location concerned by this study: Montbartier, Lectoure, Cappelle, Grisolles and Monbequi. Finally, there 7 experiments that will be useful in this study. Let’s summarize it in a table.

A=matrix("-",5,4)
rownames(A)=c("Cappelle","Grisolles","Lectoure","Monbequi","Monbartier")
colnames(A)=c("2011", "2013", "2014","2015")
v2011=c("-","Yes","-","Yes","-")
v2013=c("Yes","Yes","-","-","-")
v2014=c("-","-","Yes","-","Yes")
v2015=c("-","Yes","-","-","-")
A[,c(1:4)]=c(v2011,v2013,v2014,v2015)

print(xtable(A), type = "html", include.rownames = T , comment=FALSE)

Les pressions fongiques sont différentes d’une parcelle à l’autre: - faible: Montbartier 2014, Montbequi - Moyen: Grisolles - Fort : Cappelle

Phenotyping methods

2 types de résistance: - type 1: capacité à toucher un épi (on compte le nbr d’épi touchés) - type2: capacité à se propager dans un épi (on compte le nombre d’épillets touchés)

We have several phenotyping methods used in this study: +Fusariose: ++ resistance - note fusa? - number of spike with fusariose symptom - % of spike with fusariose symptpom - extrusion des antheres=sinon, le champi reste et se dvlp.

++ toxinogénèse=quantité de toxine que le champignon laisse sur le grain - DON=quantité de toxine, mesuré par dosage biochimique. (ultra chèr). Pas sur tous les essais du coup.

+Agronomical features: - flowering data - height - Long_col_epi? - Compacité épi –pourrait etre favorable au déplacement du champi - verse: a mark going from 0 to 8

• Metabolome: va intervenir en ralentissant la dispersion du champi, soit en limitant la quntité de toxine en surface (DON). On a pris des plantes infectées et des saines. Prélevement de graines immatures. On a pris 40 composés phénoliques potentiellement intéressant. Et on compare au cours du temps entre sain et sensibles pour voir si il y a une différence de concentration de ces composés. Ces 40 composés sont constitutifs: rien a voir avec la présence de champignon. Par contre ces composés phénol évoluent au fil du temp + sont différents entre résistant et sensibles.

See the “table” folder in the dropbox for an extensive description of available phenotypes for each trial and each phenotyping date.

NOTE

Let’s plot the histogram of every rep1 of the NOTE.
Pas de note pour CAP13.
On observe bien l’augmentation de la maladie dans le temps.
- GRI11 > GRI13 > GRI15 en terme de quantité de symptome.
- BEQ11 Très touché

par(mfrow=c(5,4), mar=c(4,3,1,1))
for(i in c( intersect(NOT, REP1), intersect(NOT, BEQ11) )   ){
  hist( data[,i]  ,  border=F, xlab="" , col=my_colors[6], main="", las=1)
  abline( v=c( data[rownames(data)=="dic2", i], data[rownames(data)=="silur", i]), col=c("blue","red") , lwd=2)
  a=gsub("_rep1", "", colnames(data)[i])
  a=paste(a, " | ", round(mean(data[,i],na.rm=T),2), sep="")
  title( a , col.main="blue" , cex.main=0.7)
  }

PEPI

Pas de PEPI pour GRI11 seulement. On a bien qqchose entre 0 et 100. Observation:
- Not a normal distribution –> importance des blups. - CAP13 / GRI13: for taux d’épi touchés. On constate une net évolution au cours du temps : hist par vers la droite - LEC14 / BAR 14: faible taux d’épi touché, moins de contamination? - Répartition plus uniforme pour les AUDPC

par(mfrow=c(6,4), mar=c(2,3,1,1))
for(i in c( intersect(PEPI, REP1), intersect(PEPI, BEQ11) )  ){
  hist( data[,i]  ,  border=F, xlab="" , col=my_colors[3], main="", las=1)
  abline( v=c( data[rownames(data)=="dic2", i], data[rownames(data)=="silur", i]), col=c("blue","red") , lwd=2)
  a=gsub("_rep1", "", colnames(data)[i])
  a=paste(a, " | ", round(mean(data[,i],na.rm=T),2), sep="")
  title( a, col.main="blue" , cex.main=0.7)
  }

NEPIL

Dispo pour toutes les parcelles.
Observation:

par(mfrow=c(7,5), mar=c(4,3,1,1))
for(i in c( intersect(NEPIL, REP1), intersect(NEPIL, BEQ11) ) ){
  hist( data[,i]  ,  border=F, xlab="" , col=my_colors[4], main="", las=1)
  abline( v=c( data[rownames(data)=="dic2", i], data[rownames(data)=="silur", i]), col=c("blue","red") , lwd=2)
  a=gsub("_rep1", "", colnames(data)[i])
  a=paste(a, " | ", round(mean(data[,i],na.rm=T),2), sep="")
  title( a, col.main="blue" , cex.main=0.7)
  }

PEPIL

par(mfrow=c(5,4), mar=c(4,3,1,1))
for(i in c( intersect(PEPIL, REP1), intersect(PEPIL, BEQ11) )   ){
  hist( data[,i]  ,  border=F, xlab="" , col=my_colors[5], main="", las=1)
  abline( v=c( data[rownames(data)=="dic2", i], data[rownames(data)=="silur", i]), col=c("blue","red") , lwd=2)
  a=gsub("_rep1", "", colnames(data)[i])
  a=paste(a, " | ", round(mean(data[,i],na.rm=T),2), sep="")
  title( a, col.main="blue" , cex.main=0.7)
  }

BLUP

Let’s plot the histogram of every BLUPs of the NOTE.
On doit voir Dic2 et Silur très très proche du coup, plus de diff entre les 2 !

par(mfrow=c(4,4), mar=c(4,3,1,1))
for(i in c( intersect(NOT, BLUP), intersect(NOT, BEQ11) )   ){
  hist( data[,i]  ,  border=F, xlab="" , col=my_colors[6], main="", las=1)
  abline( v=c( data[rownames(data)=="dic2", i], data[rownames(data)=="silur", i]), col=c("blue","red") , lwd=2)
  a=gsub("_rep1", "", colnames(data)[i])
  a=paste(a, " | ", round(mean(data[,i],na.rm=T),2), sep="")
  title( a , col.main="blue" , cex.main=0.7)
  }

Par parcelle

On va garder la rep1 seulement à chaque fois, et faire une ACP pour chaque expé:

plot_my_PCA=function( exp){
  a=intersect(exp, c(FLO, HEI, PEPI, NEPIL, PEPIL, NOT, DON))
  a=intersect(a, REP1)
  res.PCA = PCA(data[ , a ] , scale.unit=TRUE, ncp=3, graph=F) 
  plot.PCA(res.PCA, axes=c(1, 2), choix="var", cex=0.7, title="" )
}
# CAP13
plot_my_PCA(CAP13)

## Warning in PCA(data[, a], scale.unit = TRUE, ncp = 3, graph = F): Missing
## values are imputed by the mean of the variable: you should use the
## imputePCA function of the missMDA package

# GRI11
plot_my_PCA(GRI11)

## Warning in PCA(data[, a], scale.unit = TRUE, ncp = 3, graph = F): Missing
## values are imputed by the mean of the variable: you should use the
## imputePCA function of the missMDA package

# GRI13
plot_my_PCA(GRI13)

## Warning in PCA(data[, a], scale.unit = TRUE, ncp = 3, graph = F): Missing
## values are imputed by the mean of the variable: you should use the
## imputePCA function of the missMDA package

# GRI15
plot_my_PCA(GRI15)

## Warning in PCA(data[, a], scale.unit = TRUE, ncp = 3, graph = F): Missing
## values are imputed by the mean of the variable: you should use the
## imputePCA function of the missMDA package

# LEC14
plot_my_PCA(LEC14)

## Warning in PCA(data[, a], scale.unit = TRUE, ncp = 3, graph = F): Missing
## values are imputed by the mean of the variable: you should use the
## imputePCA function of the missMDA package

# BEQ11
a=intersect(BEQ11, c(FLO, HEI, PEPI, NEPIL, PEPIL, NOT, DON))
res.PCA = PCA(data[ , a ] , scale.unit=TRUE, ncp=3, graph=F) 

## Warning in PCA(data[, a], scale.unit = TRUE, ncp = 3, graph = F): Missing
## values are imputed by the mean of the variable: you should use the
## imputePCA function of the missMDA package

plot.PCA(res.PCA, axes=c(1, 2), choix="var", cex=0.7, title="" )

# BAR14
plot_my_PCA(BAR14)

## Warning in PCA(data[, a], scale.unit = TRUE, ncp = 3, graph = F): Missing
## values are imputed by the mean of the variable: you should use the
## imputePCA function of the missMDA package

Par trait

Pour voir si des expé ont tendance à se ressembler et si les individus résistants sont stables d’une parcelle à l’autre.

# Function for the PCA
replot_my_PCA=function(trait){
  a=c( intersect(trait, REP1), intersect(trait, BEQ11)) 
  b=as.numeric( as.factor( gsub("_.*","", colnames(data)[a]) ))
  res.PCA = PCA(data[ , a ] , scale.unit=TRUE, ncp=3, graph=F) 
  plot.PCA(res.PCA, axes=c(1, 2), choix="var", cex=0.7, title="" , col.var = my_colors[b])
}


# Function to calculate correlation
calculate_cor=function(trait){
  a=c( intersect(trait, REP1), intersect(trait, BEQ11)) 
  a=intersect(a, AUDPC)
  out=round(cor(data[,a] , use="complete.obs"),2)
  colnames(out)=gsub("_.*","",colnames(out))
  rownames(out)=colnames(out)
  diag(out)="-"
  out[lower.tri(out)]="-"
  return(out)
}

Apply it

## Warning in PCA(data[, a], scale.unit = TRUE, ncp = 3, graph = F): Missing
## values are imputed by the mean of the variable: you should use the
## imputePCA function of the missMDA package

## Warning in PCA(data[, a], scale.unit = TRUE, ncp = 3, graph = F): Missing
## values are imputed by the mean of the variable: you should use the
## imputePCA function of the missMDA package

## Warning in PCA(data[, a], scale.unit = TRUE, ncp = 3, graph = F): Missing
## values are imputed by the mean of the variable: you should use the
## imputePCA function of the missMDA package

## Warning in PCA(data[, a], scale.unit = TRUE, ncp = 3, graph = F): Missing
## values are imputed by the mean of the variable: you should use the
## imputePCA function of the missMDA package

Agronomical features

Let’s plot a PCA to see the relationship between variables height, precocity and flowering time. There is a strong relationship between FLO and EPI. These 2 variables are not correlated with height. These 3 variable are very stable among experiments.

res.PCA = PCA(data[ , c(EPI, FLO, HEI) ] , scale.unit=TRUE, ncp=3, graph=F) 

## Warning in PCA(data[, c(EPI, FLO, HEI)], scale.unit = TRUE, ncp = 3, graph
## = F): Missing values are imputed by the mean of the variable: you should
## use the imputePCA function of the missMDA package

ze_col=c( rep(my_colors[1], length(EPI)), rep(my_colors[2], length(FLO)), rep(my_colors[3], length(HEI)) )
plot.PCA(res.PCA, axes=c(1, 2), choix="var", cex=0.7, col.var=ze_col, title="" )

Important Blups

ACP avec les Blups des AUDPC et des dernières notations. Couleur=parcelle.

a=intersect(BLUP,c(AUDPC,LAST))
b=as.numeric( as.factor( gsub("_.*","", colnames(data)[a]) ))
res.PCA = PCA(data[ , a ] , scale.unit=TRUE, ncp=3, graph=F) 

## Warning in PCA(data[, a], scale.unit = TRUE, ncp = 3, graph = F): Missing
## values are imputed by the mean of the variable: you should use the
## imputePCA function of the missMDA package

plot.PCA(res.PCA, axes=c(1, 2), choix="var", cex=0.7, title="", col.var = my_colors[b] )

Conclusion

• FLO et EPI sont très corrélés. On peut en garder un des 2 seulement.
  
• Plantes plus haute = plante moins malade. Grosse corrélation négative. –> cafacteur pour les QTLs? Ou dans le calcul des blups?
• à l’échelle de l’expé, les variable se groupe par date d’observation, pas par méthode de phénotypage! Donc méthode de phéno = hautement correlées. Date d’observation = importante.
• Grande variance inter expé. Que ce soit variable par variable ou en prenant toutes les variables, on a systématiquement un regroupement par expé très net qui apparait.